A novel and efficient personalized stress detection technique using a deep learning model
Abstract Stress is a type of mental tension or escalated psycho-physiological state of the human body caused by a problematic situation. It majorly affects adults and elders, which further leads to chronic health problems and heart-related diseases. Various techniques were developed to detect stress...
Saved in:
| Main Authors: | , , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Nature Portfolio
2025-08-01
|
| Series: | Scientific Reports |
| Subjects: | |
| Online Access: | https://doi.org/10.1038/s41598-025-16642-w |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Abstract Stress is a type of mental tension or escalated psycho-physiological state of the human body caused by a problematic situation. It majorly affects adults and elders, which further leads to chronic health problems and heart-related diseases. Various techniques were developed to detect stress levels. Detecting the stress level through tweets is a difficult task. Several techniques were introduced to detect the stress level through text from social media using both machine learning (ML) and deep learning (DL) models. Several limitations occurred, such as higher training time, high time consumption, and limited features utilized to train the model. The proposed novelty of the technique is developed to overcome these issues and provide efficient stress detection. The novelty lies in the integration of multiple advanced text representation techniques, such as FastText, Global Vectors for Word Representation (Glove), DeepMoji, and XLNet, with Depth-wise Separable Convolution with Residual Network (DSC-ResNet) for accurate stress detection. The Chaotic Fennec Fox Optimization Algorithm (CFFO) tunes the hyperparameter. The DSC-ResNet model is improved by hybridizing the layer of depthwise separable convolution into the ResNet model. The proposed model is implemented in the Python platform. The overall performance of the proposed technique is analyzed using various related methods to determine its efficiency. The accuracy of the proposed technique is obtained to be 98.42%, which achieves higher performance compared to other associated techniques. The proposed technique’s precision, recall, specificity, and F1-score achieve 97.58%, 98.12%, 98.28%, and 98.38%. |
|---|---|
| ISSN: | 2045-2322 |